Aircraft Detection in Satellite Imagery based on Deep Learning/AI Techniques and FPGA /SoC based Hardware Implementation
Dimple Garg, Dhara Shah, Vaishnavi Chintapalli, Sandip Paul, Ashish B Mishra · 2024
Object detection in satellite imagery has become increasingly important in remote sensing surveillance applications, where the detection of aircraft is very useful. However, efficient detection of aircraft presents numerous challenges including variations in size, aspect ratio, orientation and complex background environments. Using deep learning techniques, we optimize object detection frameworks, feature extraction networks and hyperparameters for enhancing precision while minimizing detection time and achieve an AP score of 0.91. Additionally, our high speed hardware based system is optimized for near real time network deployment, achieving a notable frame rate of 46 fps on Zynq Ultrascale SOC based hardware.